An improved image processing system involves decoding compressed image data including frequency domain coefficients defining blocks of pixel values representing an image at a first resolution to provide an image at a reduced second resolution for display from a selected sub-set of the frequency domain coefficients. The apparatus includes an enhanced motion-compensation-unit (MCU) operating with blocks of pixel values representing an image at an intermediate third resolution lower than the first resolution but higher than the reduced second resolution.

Patent
   6993076
Priority
May 11 1999
Filed
Oct 28 1999
Issued
Jan 31 2006
Expiry
May 03 2021
Extension
553 days
Assg.orig
Entity
Large
19
16
all paid
1. In apparatus for decoding compressed image data including frequency domain coefficients defining blocks of pixel values representing an image at a first resolution to provide an image at a reduced second resolution for display, said apparatus comprising:
first means responsive to a selected sub-set of said frequency domain coefficients for deriving said image of said reduced second resolution for display and including,
enhanced motion-compensation-unit (MCU) processing means; and
second means for operating said enhanced MCU processing means with blocks of pixel values representing said image at an intermediate third resolution lower than said first resolution and higher than said reduced second resolution.
18. In a system for decoding compressed image data in the form of pixel blocks representing an image of a first resolution to provide an image of a reduced second resolution, a method comprising the steps of:
generating data representative of an image pixel block at an intermediate third resolution lower than said first resolution but higher than said reduced second resolution;
generating motion compensated pixel block data at said third resolution from said pixel block data of said reduced second resolution supplemented by said intermediate third resolution data; and
deriving pixel data representing said image of said reduced second resolution from said motion compensated pixel block data at said third resolution.
11. In a system for decoding compressed image data in the form of pixel blocks representing an image at a first resolution to provide an image of a reduced second resolution, a method of decompressing a pixel block of said first resolution by:
selecting a sub-set of frequency domain coefficients in said pixel blocks of said compressed image data;
processing elements of said sub-set of frequency domain coefficients to provide pixel data representing pixels comprising a spatially distributed sub-set of pixels in a pixel block of said image at a first resolution and excluding other pixels of that pixel block, said processing including
using data at an intermediate third resolution, lower than said first resolution but higher than said reduced second resolution, to supplement data from said reduced second resolution in forming prediction for motion compensation; and
formatting said pixel data representing pixels comprising said spatially distributed sub-set of pixels to provide said image of said reduced second resolution.
5. In apparatus for decoding compressed image data including frequency domain coefficients defining blocks of pixel values representing an image at a first resolution to provide an image at a reduced second resolution for display, said apparatus comprising:
first means responsive to a selected sub-set of said frequency domain coefficients for deriving said image of said reduced second resolution for display and including,
enhanced motion-compensation-unit (MCU) processing means; and
second means for operating said enhanced MCU processing means with blocks of pixel values representing said image at an intermediate third resolution lower than said first resolution and higher than said reduced second resolution, wherein
said enhanced MCU processing means is responsive to base-layer pixel macroblock input values representing said image at said reduced second resolution and to pixel values representing said image at said intermediate third resolution for deriving motion-compensated base-layer prediction macroblock output pixel values as a first output and motion-compensated enhancement-layer prediction macroblock output pixel residual values as a second output.
2. The apparatus defined in claim 1, wherein
said reduced second resolution is substantially ¼ of said first resolution; and
said second means operates said enhanced MCU processing at an intermediate third resolution which is substantially ½ of said first resolution.
3. The apparatus defined in claim 1, wherein
said image at said reduced second resolution for display is a progressive-scanned image.
4. The apparatus defined in claim 1, wherein
said image at said reduced second resolution for display is an interlaced-scanned image.
6. The apparatus defined in claim 5 wherein
said second means comprises third means responsive to said selected sub-set of said frequency domain coefficients and to both said motion-compensated base-layer macroblock output pixel values and said enhancement-layer macroblock output pixel residual values for deriving both said base-layer macroblock input pixel values and said encoded enhancement-layer macroblock input pixel residual values.
7. The apparatus defined in claim 5, wherein said second means comprises:
a base and enhancement-layer decimated-pixel memory;
unitary enhanced inverse discrete cosine transform (DCT), filtering and pixel-decimation processing means responsive to a selected sub-set of frequency domain coefficients for deriving base-layer blocks of output pixel values representing said image at said reduced second resolution as a first output and output enhancement-layer blocks of output pixel residual values representing said image at said intermediate third resolution as a second output;
fourth means, including a first adder for adding corresponding pixel values of said motion-compensated base-layer macroblock output pixel values from said enhanced MCU processing means and said base-layer blocks of output pixel values from said unitary IDCT, filtering and pixel-decimation processing means, for deriving values that are stored as base-layer data in said base and enhancement-layer decimated-pixel memory;
fifth means, including a second adder and an enhancement-layer encoder, for adding corresponding pixel residual values of said motion-compensated enhancement-layer macroblock output pixel residual values from said enhanced MCU processing means to said enhancement-layer blocks of output pixel residual values from said unitary IDCT, filtering and pixel-decimation processing means to obtain a sum output from said second adder for encoding by said enhancement-layer encoder, for deriving second input values that are stored as encoded enhancement-layer data in said base and enhancement-layer decimated-pixel memory; and
sixth means for providing from said base and enhancement-layer decimated-pixel memory said base-layer pixel macroblock input values to said enhanced MCU processing means and for deriving said encoded enhancement-layer pixel macroblock input residual values applied as a second input to said enhanced MCU processing means from said stored encoded enhancement-layer data.
8. The apparatus defined in claim 5, wherein
said frequency domain coefficients define image information that includes luma blocks of pixel values representing intracoded (I) and predictive-coded (P) progressive-scanned image at said first resolution.
9. The apparatus defined in claim 7, including
seventh means comprising a sample-rate converter for deriving an ongoing display video signal from base-layer blocks of output pixel values.
10. The apparatus defined in claim 1, wherein
said reduced second resolution is substantially ¼ of said first resolution; and
said intermediate third resolution is substantially ½ of said first resolution.
12. A method according to claim 11 including the step of
selecting different spatially distributed sub-sets of pixels for interlace and progressive image output.
13. A method according to claim 11 wherein said formatting step comprises,
upsampling said pixel data representing pixels comprising a spatially distributed sub-set of pixels to provide said image of said reduced second resolution.
14. A method according to claim 11 wherein said processing step includes the step of,
Selecting said spatially distributed sub-set of pixels based on desired PIP picture characteristic.
15. A method according to claim 14 wherein
said PIP picture characteristic comprises at least one of (a) PIP picture size, (b) whether said PIP picture is interlace or progressive, and (c) PIP picture vertical and horizontal pixel resolution.
16. A method according to claim 11 wherein said formatting step includes the step of,
adaptively filtering pixel data representing pixels comprising a spatially distributed sub-set of pixels using a filter function selected based on at least one of, (a) motion vector type, (b) group of picture (GOP) structure, (c) a GOP boundary transition, (d) whether I, B or P frame, and (e) whether interlace or progressive frame reduced second resolution output required.
17. A method according to claim 11 wherein said formatting step includes the step of,
adaptively filtering pixel data representing pixels comprising a selected spatially distributed sub-set of pixels using a filter function selected from at least one of, (a) a vertical pixel data filter, (b) a horizontal pixel data filter, (c) a chrominance data filter, and (d) luminance data filter.
19. A method according to claim 18 wherein the steps of claim 18 are performed for P frames exclusively of I and B frames.
20. A method according to claim 18 wherein the steps of claim 18 are performed for P frames and one of, (a) I frames and (b) B frames.
21. A method according to claim 18 including the step of
upsampling said pixel block data at said third resolution to provide image data of said first resolution.
22. A method according to claim 21 including the step of
downsampling said upsampled pixel block data of said first resolution to provide image data of said second resolution.
23. A method according to claim 21 including the step of
downsampling said upsampled pixel block data of said first resolution to provide said intermediate third resolution data.
24. A method according to claim 18 wherein
said pixel block data of said third resolution comprises residual data.

This is a non-provisional application of provisional application Ser. No. 60/133,429 by M. L. Corner et al, filed 11 May 1999.

The present invention relates to the decoding of a coded high-definition (HD) video signal to derive an enhanced decoded video signal suitable, for example, for recording or producing a picture-in-picture (PIP) or other reduced-resolution display.

Known in the art are television receivers that, while displaying a relatively large picture derived from a primary television channel, also simultaneously display a small picture-in-picture (PIP) derived from a secondary television channel. In the case of a high-definition television (HDTV) receiver, the receiver must include a relatively complex and expensive decoder that conforms with the MPEG ISO 13818-2 standard for decoding a received coded HD video signal in real time for high definition display. However, because the PIP is small, there is no need to provide a high definition PIP display because a viewer inherently would not be able to resolve the higher definition components of a high definition PIP. Therefore, to provide the PIP, the HDTV receiver may be supplied with a lower-resolution second simpler and less expensive decoder which still conforms with the ISO 13818-2 standard.

One approach, known in the art, to providing a lower-resolution second decoder which is somewhat simpler and less expensive than the decoder providing the high definition display, is disclosed in the three U.S. Pat. Nos. 5,614,952, 5,614,957 and 5,635,985, which were, respectively, issued to Boyce et al. on Mar. 25, 1997, Mar. 25, 1997 and Jun. 3, 1997.

Further, incorporated herein by reference is the teaching of copending U.S. patent application Ser. No. 09/349,865, filed Jul. 8, 1999 and assigned to the same assignee as the present application, which is directed to a lower-resolution second-decoder approach suitable for deriving a PIP display in real time from a received coded HD video signal that is significantly simpler and less expensive to implement than is the second decoder disclosed by Boyce et al, but still conforms with the ISO 13818-2 standard.

A system involves decoding compressed image data including frequency domain coefficients defining blocks of pixel values representing an image at a first resolution to provide an image at a reduced second resolution. The system includes a motion-compensation-unit (MCU) processor responsive to a selected sub-set of the frequency domain coefficients for deriving the image of the reduced second resolution. The motion-compensation-unit (MCU) processor employs blocks of pixel values representing image data at an intermediate third resolution lower than the first resolution and higher than the reduced second resolution.

FIG. 1 is a functional block diagram showing a variable-length decoder (VLD) responsive to an input HD MPEG data bit-stream for providing a first selected MPEG data output to a PIP decoding means and a second selected MPEG data output to an HD decoding means;

FIG. 1a shows an 8×8 block containing the 64 DCT coefficients that are employed by the HD decoding means of FIG. 1, FIG. 1b shows an 8×8 block containing the particular 10 DCT coefficients of the 64 DCT coefficients shown in FIG. 1a that are employed by the PIP decoding means of FIG. 1 for progressive-scan sequences and FIG. 1c shows an 8×8 block containing the particular 10 DCT coefficients of the 64 DCT coefficients shown in FIG. 1a that are employed by the PIP decoding means of FIG. 1 for interlaced-scan sequences;

FIG. 2 is a simplified functional block diagram of an embodiment of the PIP decoding means of FIG. 1 which incorporates features of the present invention;

FIG. 3 is a functional block diagram showing details of the enhanced MCU processing means of FIG. 2.

FIG. 4 is a conceptual diagram showing the computational processing performed by the DCT-based upsample means of FIG. 3; and

FIG. 5 is a conceptual diagram showing the computational processing performed by the DCT-based downsample means of FIG. 3.

Referring to FIG. 1, there is shown VLD 100, PIP decoding means 102 and HD decoding means 104. In accordance with the known teaching of the MPEG ISO 13818-2 standard, one of the responses of VLD 100 to the input coded HD MPEG data comprising a sequence of MPEG I, P and B frames is to convey coded picture information defined by each of successive 8×8 blocks of quantized discrete cosine transform (DCT) coefficients as an input to HD decoding means 104. Further, in accordance with the known teaching of the MPEG ISO 13818-2 standard, among the functions performed by HD decoding means 104 is to first perform inverse quantization of each successive 8×8 block of DCT coefficients and then perform inverse discrete cosine transformation (IDCT) of the DCT coefficients of each successive 8×8 block. Finally, HD decoding means 104 must perform motion compensation for each P frame and bi-directionally predictive B frame after IDCT has been performed on that P or B frame.

FIG. 1a shows an 8×8 block of DCT coefficients, wherein (1) the value of coefficient DCT0,0 (located in the upper left corner of the 8×8 block) represents the average (DC) value (i.e., both the horizontal and vertical frequencies are 0) of the picture defined by the 64 values of a corresponding 8×8 block of pixels prior to having undergone DCT, while (2) the value of coefficient DCT7,7 (located in the lower right corner of the 8×8 block) represents the highest horizontal frequency and highest vertical frequency components of the picture defined by the 64 values of a corresponding 8×8 block of pixels prior to having undergone DCT. For the case of a HD picture, all, or nearly all, of the 64 DCT coefficients from DCT0,0 to DCT7,7 inclusive of FIG. 1a may have non-zero values. This results in a relatively large amount of image-processing computation to accomplish IDCT in real time. Further, motion compensation also involves a large amount of real time image-processing computation. Therefore, HD decoding means 104 requires about 96 Mbits memory to temporarily store MPEG decoded image frames prior to display. HD decoding means 104 requires these frames for motion compensation to reconstruct accurate images for display. Thus, a physical implementation of HD decoding means 104 is relatively expensive.

Returning to FIG. 1, another of the responses of VLD 100 to the input coded HD MPEG data is to convey only coded picture information defined by a relatively small given number of lower-frequency-defining, quantized DCT coefficients of each successive 8×8 block as an input to PIP decoding means 102. It is to be noted that the PIP processing and images and the term PIP itself is used herein to encompass any form of reduced resolution image and processing and not just television PIP image generation. While the preferred tutorial example of the PIP decoding means described in the aforesaid patent application Ser. No. 09/349,865 employed only the 6 lowest-frequency quantized DCT coefficients, the preferred tutorial example of enhanced-quality PIP decoding means 102, described in detail below, employs 10 DCT coefficients consisting of DCT0,0, DCT1,0, DCT2,0, DCT3,0, DCT0,1, DCT1,1, DCT2,1, DCT0,2, DCT1,2, and DCT0,3 shown in FIG. 1b for progressive-scan use or, alternatively, consisting of DCT0,0, DCT1,0, DCT2,0, DCT0,1, DCT1,1, DCT0,2, DCT0,3, DCT0,4, DCT0,5, and DCT0,6 shown in FIG. 1c for interlaced-scan use, thereby providing better high-frequency response for the enhanced PIP display. More specifically, the PIP bit-stream received by VLD 100 has been pre-parsed by a VLD-PIP parser (not shown to simplify FIG. 1) to remove from the bitstream DCT coefficients which are not needed by the PIP decoder.

The simplified functional block diagram of the embodiment of enhanced-quality PIP decoding means 102 shown in FIG. 2 comprises runlength decoder (RLD) 200, inverse quantizer (IQ) 202, unitary enhanced IDCT, filtering and pixel-decimation processing means 204, base-layer adder 205B, enhancement-layer adder 205E, base and enhancement-layer decimated pixel memory 206, enhancement-layer encoder 207, enhanced motion compensation unit (MCU) processing means 208 and sample-rate converter 210. Although the simplified functional block diagram of FIG. 2 does not show means for controlling the operation of this embodiment of enhanced-quality PIP decoding means 102, it should be understood that suitable control means that conform with the requirements of the ISO 13818-2 standard is included in a physical implementation of this embodiment.

For illustrative purposes, the following description of elements 200, 202, 204, 205B, 205E, 206, 207, 208 and 210 assumes that each of these elements is being operated in accordance with the above-discussed preferred tutorial example.

In this example, RLD 200 outputs 10 DCT coefficients for each 8×8 coded block using the 2 scan patterns defined in the ISO 13818-2 standard. The positioning of the 10 DCT coefficients within each 8×8 block is illustrated in FIG. 1b for progressive scan and in FIG. 1c for interlaced scan, as determined by the state of the progressivesequence flag. In the FIG. 1b progressive sequence case, if the alternatescan flag from the picture coding extension for the current picture is 0, the 10 DCT coefficients correspond to coefficients 0,1,2,3,4,5,6,7,8,9 in 1-dimensional scan order, whereas if alternatescan is 1, the 10 DCT coefficients of interest are coefficients 0,1,2,3,4,5,6,7,8,20 in scan order. In the FIG. 1c interlaced sequence case, if the alternatescan flag from the picture coding extension for the current picture is 0, the 10 DCT coefficients correspond to coefficients 0,1,2,3,4,5,9,10,20,21 in 1-dimensional scan order, whereas if alternatescan is 1, the 10 DCT coefficients of interest are coefficients 0,1,2,3,4,5,6,10,11,12 in scan order. There are two run values that have a meaning in RLD 200 which is different from that described in the ISO 13818-2 standard, depending on the value of the alternatescan and progressivesequence flags. For progressive sequences, if alternatescan is 0, a run value of 10 indicates that the coefficients needed by PIP decoder 102 are all 0 and there is no subsequent non-zero coefficient. Similarly, if alternatescan is 1, a run value of 21 indicates that the coefficients needed by decoder 102 are all 0 and there is no subsequent non-zero coefficient. For interlaced sequences, if alternatescan is 0, a run value of 22 indicates that the coefficients needed by PIP decoder 102 are all 0 and there is no subsequent non-zero coefficient. Similarly, if alternatescan is 1, a run value of 13 indicates that the coefficients needed by decoder 102 are all 0 and there is no subsequent non-zero coefficient. Table 1, summarizes the meaning of run values of 10 and 21 for the two possible values of the alternatescan flag for progressive sequence and Table 2, summarizes the meaning of run values of 13 and 22 for the two possible values of the alternatescan flag for interlaced sequence. All other alternatescan/run value combinations encountered by RLD 200 are interpreted as described in the ISO 13818-2 standard.

TABLE 1
Interpretation of run = 10 and run = 21 by RLD 200
for progressive sequences
Run AlternateScan Interpretation in PIP RLD
10 0 All DCT coefficients = 0
10 1 Same as ISO 13818-2 standard
21 0 Not allowed
21 1 All DCT coefficients = 0

TABLE 2
Interpretation of run = 13 and run = 22 by RLD 200
for interlaced sequences
Run AlternateScan Interpretation in PIP RLD
13 0 Same as ISO 13818-2 standard
13 1 All DCT coefficients = 0
22 0 All DCT coefficients = 0
22 1 Not allowed

IQ 202 performs inverse quantization arithmetic and saturation described in the ISO 13818-2 standard on the 10 DCT coefficients shown in FIG. 1b for progressive sequences and shown in FIG. 1c for interlaced sequences. The mismatch control portion of the inverse quantization process is not needed. Conventionally, an extensive computational process requiring three separate steps is needed to convert the coded frequency domain information in an 8×8 block at the output from IQ 202, into spatial domain picture information comprising respective values of a smaller block of decimated pixels of a reduced-resolution PIP display image. The first step is to determine the value of each of the 64 (i.e., full pixel density) pixel values of each 8×8 block of picture information as an IDCT function of the inversely-quantized DCT coefficient values. Thereafter, the second step of lowpass filtering followed by the third step of pixel decimation may be performed on the pixels in each successive 8×8 block to provide the desired smaller block of decimated pixels. For instance, a decimation of alternate horizontal and vertical filtered pixels for the case of a progressive scan would result in a 75% reduction in pixel density. Similarly, a decimation of 3 out of 4 successive filtered pixels in the horizontal direction for the case of an interlaced scan would also result in a 75% reduction in pixel density. Thus, in either case, such a decimation performed for luma pixels and also for chroma pixels would result in a reduction in pixel density from 64 per 8×8 block to only 16 per 8×8 block for each of them. However, the amount of hardware required to implement this conventional three-step computational process is relatively large and, therefore, relatively expensive.

In accordance with the principles of the invention taught in the aforesaid patent application Ser. No. 09/349,865, the unitary IDCT, filtering and pixel-decimation processing means disclosed therein is able to convert the coded respective values of inversely-quantized DCT coefficients contained in an 8×8 block at the output from IQ 202, into a smaller block of decimated pixels in a single-step computational process. Thus, the amount of hardware required to implement this single-step computational process by means 204 is relatively small and, therefore, relatively inexpensive compared to the aforesaid conventional three-step computational process.

Specifically, in accordance with the teachings of the aforesaid patent application Ser. No. 09/349,865, the decimated pixel memory thereof (which, because of pixel decimation, requires a storage capacity size of only ¼ the capacity size of a corresponding undecimated pixel memory) comprises a plurality of separate buffers. Each of these buffers is capable of temporarily storing decimated luma and chroma pixels. In conformity with the ISO 13818-2 standard, the decimated pixel memory includes one or more buffers for storing decimated pixels that define reconstructed intracoded (I), predictive-coded (P) and/or bi-directionally predictive-coded (B) frame or field pictures. Further, motion-compensated prediction macroblock output of pixel values from the MCU processing means is added in an adder to each corresponding macroblock output derived in the unitary IDCT, filtering and pixel-decimation processing means. The summed pixel values of the output from the adder are stored into a first buffer of the decimated pixel memory. This first buffer may be a first-in first-out (FIFO) buffer in which the stored decimated pixels may be reordered between (1) being written into the first buffer and (2) being read out from the first buffer and written into another buffer of the decimated pixel memory. In the case of a current P or B frame or field, the decimated pixel memory includes a buffer for storing a macroblock for input to the MCU processing means to provide motion compensation.

In accordance with the principles of the present invention, two layers of decimated pixels are advantageously stored respectively in base and enhancement-layer decimated pixel memory 206 to achieve high-quality motion compensation and improve PIP image quality. The first of these two layers is a base-layer of decimated pixels and the second of these two layers is an enhancement-layer of vector-quantized values of luma macroblock decimated pixels that are employed in enhanced MCU processing means 208 during decoding of P pictures The enhanced layer is used to provide a reduced resolution image of greater resolution than is obtainable by using just the decimated pixels of the base-layer. Both the base-layer and this enhancement-layer are employed by enhanced MCU processing means 208, in a manner described in detail below.

Preferred embodiments of IDCT, filtering and pixel-decimation processing means 204, enhancement-layer encoder 207 and enhanced MCU processing means 208 for implementing the present invention will now be described in detail.

The unitary enhanced IDCT, filtering and pixel-decimation processing means 204 provides the following sets of 16 decimated pixel values for each of the base and enhancement-layers used for each of progressive scan and interlaced scan (each set being a function of 10 DCT coefficient values).
Progressive-Scan, Base-Layer Set of Decimated Pixel Values g 1 ( 0 , 0 ) = [ 8 DCT 0 , 0 + 10 DCT 1 , 0 + 7 DCT 2 , 0 + 4 DCT 3 , 0 + 10 DCT 0 , 1 + 13 DCT 1 , 1 + 9 DCT 2 , 1 + 7 DCT 0 , 2 + 9 DCT 1 , 2 + 4 DCT 0 , 3 ] / 64 g 1 ( 1 , 0 ) = [ 8 DCT 0 , 0 + 4 DCT 1 , 0 - 7 DCT 2 , 0 - 9 DCT 3 , 0 + 10 DCT 0 , 1 + 5 DCT 1 , 1 - 9 DCT 2 , 1 + 7 DCT 0 , 2 + 4 DCT 1 , 2 + 4 DCT 0 , 3 ] / 64 g 1 ( 2 , 0 ) = [ 8 DCT 0 , 0 - 4 DCT 1 , 0 - 7 DCT 2 , 0 + 9 DCT 3 , 0 + 10 DCT 0 , 1 - 5 DCT 1 , 1 - 9 DCT 2 , 1 + 7 DCT 0 , 2 - 4 DCT 1 , 2 + 4 DCT 0 , 3 ] / 64 g 1 ( 3 , 0 ) = [ 8 DCT 0 , 0 - 10 DCT 1 , 0 + 7 DCT 2 , 0 - 4 DCT 3 , 0 + 10 DCT 0 , 1 - 13 DCT 1 , 1 + 9 DCT 2 , 1 + 7 DCT 0 , 2 - 9 DCT 1 , 2 + 4 DCT 0 , 3 ] / 64 g 1 ( 0 , 1 ) = [ 8 DCT 0 , 0 + 10 DCT 1 , 0 + 7 DCT 2 , 0 + 4 DCT 3 , 0 + 4 DCT 0 , 1 + 5 DCT 1 , 1 + 4 DCT 2 , 1 - 7 DCT 0 , 2 - 9 DCT 1 , 2 - 9 DCT 0 , 3 ] / 64 g 1 ( 1 , 1 ) = [ 8 DCT 0 , 0 + 4 DCT 1 , 0 - 7 DCT 2 , 0 - 9 DCT 3 , 0 + 4 DCT 0 , 1 + 2 DCT 1 , 1 - 4 DCT 2 , 1 - 7 DCT 0 , 2 - 4 DCT 1 , 2 - 9 DCT 0 , 3 ] / 64 g 1 ( 2 , 1 ) = [ 8 DCT 0 , 0 - 4 DCT 1 , 0 - 7 DCT 2 , 0 + 9 DCT 3 , 0 + 4 DCT 0 , 1 - 2 DCT 1 , 1 - 4 DCT 2 , 1 - 7 DCT 0 , 2 + 4 DCT 1 , 2 - 9 DCT 0 , 3 ] / 64 g 1 ( 3 , 1 ) = [ 8 DCT 0 , 0 - 10 DCT 1 , 0 + 7 DCT 2 , 0 - 4 DCT 3 , 0 + 4 DCT 0 , 1 - 5 DCT 1 , 1 + 4 DCT 2 , 1 - 7 DCT 0 , 2 + 9 DCT 1 , 2 - 9 DCT 0 , 3 ] / 64 g 1 ( 0 , 2 ) = [ 8 DCT 0 , 0 + 10 DCT 1 , 0 + 7 DCT 2 , 0 + 4 DCT 3 , 0 - 4 DCT 0 , 1 - 5 DCT 1 , 1 - 4 DCT 2 , 1 - 7 DCT 0 , 2 - 9 DCT 1 , 2 + 9 DCT 0 , 3 ] / 64 g 1 ( 1 , 2 ) = [ 8 DCT 0 , 0 + 4 DCT 1 , 0 - 7 DCT 2 , 0 - 9 DCT 3 , 0 - 4 DCT 0 , 1 - 2 DCT 1 , 1 + 4 DCT 2 , 1 - 7 DCT 0 , 2 - 4 DCT 1 , 2 + 9 DCT 0 , 3 ] / 64 g 1 ( 2 , 2 ) = [ 8 DCT 0 , 0 - 4 DCT 1 , 0 - 7 DCT 2 , 0 + 9 DCT 3 , 0 - 4 DCT 0 , 1 + 2 DCT 1 , 1 + 4 DCT 2 , 1 - 7 DCT 0 , 2 + 4 DCT 1 , 2 + 9 DCT 0 , 3 ] / 64 g 1 ( 3 , 2 ) = [ 8 DCT 0 , 0 - 10 DCT 1 , 0 + 7 DCT 2 , 0 - 4 DCT 3 , 0 - 4 DCT 0 , 1 + 5 DCT 1 , 1 - 4 DCT 2 , 1 - 7 DCT 0 , 2 + 9 DCT 1 , 2 + 9 DCT 0 , 3 ] / 64 g 1 ( 0 , 3 ) = [ 8 DCT 0 , 0 + 10 DCT 1 , 0 + 7 DCT 2 , 0 + 4 DCT 3 , 0 - 10 DCT 0 , 1 - 13 DCT 1 , 1 - 9 DCT 2 , 1 + 7 DCT 0 , 2 + 9 DCT 1 , 2 - 4 DCT 0 , 3 ] / 64 g 1 ( 1 , 3 ) = [ 8 DCT 0 , 0 + 4 DCT 1 , 0 - 7 DCT 2 , 0 - 9 DCT 3 , 0 - 10 DCT 0 , 1 - 5 DCT 1 , 1 + 9 DCT 2 , 1 + 7 DCT 0 , 2 + 4 DCT 1 , 2 - 4 DCT 0 , 3 ] / 64 g 1 ( 2 , 3 ) = [ 8 DCT 0 , 0 - 4 DCT 1 , 0 - 7 DCT 2 , 0 + 9 DCT 3 , 0 - 10 DCT 0 , 1 + 5 DCT 1 , 1 + 9 DCT 2 , 1 + 7 DCT 0 , 2 - 4 DCT 1 , 2 - 4 DCT 0 , 3 ] / 64 g 1 ( 3 , 3 ) = [ 8 DCT 0 , 0 - 10 DCT 1 , 0 + 7 DCT 2 , 0 - 4 DCT 3 , 0 - 10 DCT 0 , 1 + 13 DCT 1 , 1 - 9 DCT 2 , 1 + 7 DCT 0 , 2 - 9 DCT 1 , 2 - 4 DCT 0 , 3 ] / 64
Progressive-Scan, Enhancement-Layer Set of Decimated PIXEL Values g 0 ( 0 , 0 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 + 1 DCT 1 , 1 + 1 DCT 2 , 1 + 3 DCT 0 , 2 + 4 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 1 , 0 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 + 0 DCT 1 , 1 - 1 DCT 2 , 1 + 3 DCT 0 , 2 + 2 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 2 , 0 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 + 0 DCT 1 , 1 - 1 DCT 2 , 1 + 3 DCT 0 , 2 - 2 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 3 , 0 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 - 1 DCT 1 , 1 + 1 DCT 2 , 1 + 3 DCT 0 , 2 - 4 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 0 , 2 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 + 3 DCT 1 , 1 + 2 DCT 2 , 1 + 3 DCT 0 , 2 + 4 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 1 , 2 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 + 1 DCT 1 , 1 - 2 DCT 2 , 1 + 3 DCT 0 , 2 + 2 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 2 , 2 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 - 1 DCT 1 , 1 - 2 DCT 2 , 1 + 3 DCT 0 , 2 - 2 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 3 , 2 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 - 3 DCT 1 , 1 + 2 DCT 2 , 1 + 3 DCT 0 , 2 - 4 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 0 , 4 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 + 3 DCT 1 , 1 + 2 DCT 2 , 1 - 3 DCT 0 , 2 - 4 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 1 , 4 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 + 1 DCT 1 , 1 - 2 DCT 2 , 1 - 3 DCT 0 , 2 - 2 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 2 , 4 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 - 1 DCT 1 , 1 - 2 DCT 2 , 1 - 3 DCT 0 , 2 + 2 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 3 , 4 ) = 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 2 DCT 0 , 1 - 3 DCT 1 , 1 + 2 DCT 2 , 1 - 3 DCT 0 , 2 + 4 DCT 1 , 2 - 2 DCT 0 , 3 ] / 64 g 0 ( 0 , 6 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 + 1 DCT 1 , 1 + 1 DCT 2 , 1 - 3 DCT 0 , 2 + 49 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 1 , 6 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 + 0 DCT 1 , 1 - 1 DCT 2 , 1 - 3 DCT 0 , 2 - 2 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 2 , 6 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 + 0 DCT 1 , 1 - 1 DCT 2 , 1 - 3 DCT 0 , 2 + 2 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64 g 0 ( 3 , 6 ) = [ 0 DCT 0 , 0 + 0 DCT 1 , 0 + 0 DCT 2 , 0 + 0 DCT 3 , 0 + 1 DCT 0 , 1 - 1 DCT 1 , 1 + 1 DCT 2 , 1 - 3 DCT 0 , 2 + 4 DCT 1 , 2 + 6 DCT 0 , 3 ] / 64
Interlaced-Scan, Base-Layer Set of Decimated Pixel Values g 1 ( 0 , 0 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 + 11 DCT 0 , 1 + 10 DCT 1 , 1 + 10 DCT 0 , 2 + 0 DCT 2 , 0 + 9 DCT 0 , 3 + 8 DCT 0 , 4 + 6 DCT 0 , 5 + 4 DCT 0 , 6 ] / 64 g 1 ( 1 , 0 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 + 11 DCT 0 , 1 - 10 DCT 1 , 1 + 10 DCT 0 , 2 + 0 DCT 2 , 0 + 9 DCT 0 , 3 + 8 DCT 0 , 4 + 6 DCT 0 , 5 + 4 DCT 0 , 6 ] / 64 g 1 ( 0 , 1 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 + 9 DCT 0 , 1 + 9 DCT 1 , 1 + 4 DCT 0 , 2 + 0 DCT 2 , 0 - 2 DCT 0 , 3 - 8 DCT 0 , 4 - 11 DCT 0 , 5 - 10 DCT 0 , 6 ] / 64 g 1 ( 1 , 1 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 + 9 DCT 0 , 1 - 9 DCT 1 , 1 + 4 DCT 0 , 2 + 0 DCT 2 , 0 - 2 DCT 0 , 3 - 8 DCT 0 , 4 - 11 DCT 0 , 5 - 10 DCT 0 , 6 ] / 64 g 1 ( 0 , 2 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 + 6 DCT 0 , 1 + 6 DCT 1 , 1 - 4 DCT 0 , 2 + 0 DCT 2 , 0 - 11 DCT 0 , 3 - 8 DCT 0 , 4 + 2 DCT 0 , 5 + 10 DCT 0 , 6 ] / 64 g 1 ( 1 , 2 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 + 6 DCT 0 , 1 - 6 DCT 1 , 1 - 4 DCT 0 , 2 + 0 DCT 2 , 0 - 11 DCT 0 , 3 - 8 DCT 0 , 4 + 2 DCT 0 , 5 + 10 DCT 0 , 6 ] / 64 g 1 ( 0 , 3 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 + 2 DCT 0 , 1 + 2 DCT 1 , 1 - 10 DCT 0 , 2 + 0 DCT 2 , 0 - 6 DCT 0 , 3 - 8 DCT 0 , 4 + 9 DCT 0 , 5 - 4 DCT 0 , 6 ] / 64 g 1 ( 1 , 3 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 + 2 DCT 0 , 1 + 2 DCT 1 , 1 - 10 DCT 0 , 2 + 0 DCT 2 , 0 - 6 DCT 0 , 3 + 8 DCT 0 , 4 + 9 DCT 0 , 5 - 4 DCT 0 , 6 ] / 64 g 1 ( 0 , 4 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 - 2 DCT 0 , 1 - 2 DCT 1 , 1 - 10 DCT 0 , 2 + 0 DCT 2 , 0 + 6 DCT 0 , 3 + 8 DCT 0 , 4 - 9 DCT 0 , 5 - 4 DCT 0 , 6 ] / 64 g 1 ( 1 , 4 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 - 2 DCT 0 , 1 + 2 DCT 1 , 1 - 10 DCT 0 , 2 + 0 DCT 2 , 0 + 6 DCT 0 , 3 + 8 DCT 0 , 4 - 9 DCT 0 , 5 - 4 DCT 0 , 6 ] / 64 g 1 ( 0 , 5 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 - 6 DCT 0 , 1 - 6 DCT 1 , 1 - 4 DCT 0 , 2 + 0 DCT 2 , 0 + 11 DCT 0 , 3 - 8 DCT 0 , 4 - 2 DCT 0 , 5 + 10 DCT 0 , 6 ] / 64 g 1 ( 1 , 5 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 - 6 DCT 0 , 1 + 6 DCT 1 , 1 - 4 DCT 0 , 2 + 0 DCT 2 , 0 + 11 DCT 0 , 3 - 8 DCT 0 , 4 - 2 DCT 0 , 5 + 10 DCT 0 , 6 ] / 64 g 1 ( 0 , 6 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 - 9 DCT 0 , 1 - 9 DCT 1 , 1 + 4 DCT 0 , 2 + 0 DCT 2 , 0 + 2 DCT 0 , 3 - 8 DCT 0 , 4 + 11 DCT 0 , 5 - 10 DCT 0 , 6 ] / 64 g 1 ( 1 , 6 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 - 9 DCT 0 , 1 + 9 DCT 1 , 1 + 4 DCT 0 , 2 + 0 DCT 2 , 0 + 2 DCT 0 , 3 - 8 DCT 0 , 4 + 11 DCT 0 , 5 - 10 DCT 0 , 6 ] / 64 g 1 ( 0 , 7 ) = [ 8 DCT 0 , 0 + 7 DCT 1 , 0 - 11 DCT 0 , 1 - 10 DCT 1 , 1 + 10 DCT 0 , 2 + 0 DCT 2 , 0 - 9 DCT 0 , 3 + 8 DCT 0 , 4 - 6 DCT 0 , 5 + 4 DCT 0 , 6 ] / 64 g 1 ( 1 , 7 ) = [ 8 DCT 0 , 0 - 7 DCT 1 , 0 - 11 DCT 0 , 1 + 10 DCT 1 , 1 + 10 DCT 0 , 2 + 0 DCT 2 , 0 - 9 DCT 0 , 3 + 8 DCT 0 , 4 - 6 DCT 0 , 5 + 4 DCT 0 , 6 ] / 64
Interlaced-Scan, Enhancement-Layer Set of Decimated Pixel Values g 0 ( 0 , 0 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 4 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 0 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 4 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 1 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 4 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 1 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 4 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 2 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 2 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 2 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 2 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 3 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 1 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 3 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 + 1 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 4 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 1 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 4 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 1 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 5 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 2 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 5 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 2 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 6 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 4 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 6 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 4 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 0 , 7 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 4 DCT 1 , 1 + 0 DCT 0 , 2 + 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64 g 0 ( 2 , 7 ) = [ 0 DCT 0 , 0 + 3 DCT 1 , 0 + 0 DCT 0 , 1 - 4 DCT 1 , 1 + 0 DCT 0 , 2 - 7 DCT 2 , 0 + 0 DCT 0 , 3 + 0 DCT 0 , 4 + 0 DCT 0 , 5 + 0 DCT 0 , 6 ] / 64

Each of the above “Progressive-Scan Set of Decimated Pixel Values” and above “Interlaced-Scan Set of Decimated Pixel Values” was derived in the following manner:

In the case of an interlaced scan (i.e., the progressivesequence flag is 0), the base-layer value g1′(x,y) is computed in accordance with the following equation 6 and the enhancement-layer value g0′(x,y) is computed in accordance with the following equation 7: g 1 ( x , y ) = 1 4 [ f ( 4 x , y ) + f ( 4 x + 1 , y ) + f ( 4 x + 2 , y ) + f ( 4 x + 3 , y ) ] for x = 0 , 1 ; y = 0 , , 7. ( 6 ) g 0 ( x , y ) = 1 4 [ f ( 2 x , y ) + f ( 2 x + 1 , y ) - f ( 2 x + 2 , y ) - f ( 2 x + 3 , y ) ] for x = 0 , 2 ; y = 0 , , 7. ( 7 )
In the interlaced-scan case of an 8×8 block, g1′(x,y) in equation 6 defines the average value of the values of a set of 4 contiguous pixels (or prediction errors) arranged in a 4×1 block portion of the 8×8 block. The value g0′(x,y) in equation 7 defines the difference between the average value of the values of a first set of 2 contiguous horizontal pixels (or prediction errors) of a vertical line and the average value of the values of a second set of the next 2 contiguous horizontal pixels (or prediction errors) of the same vertical line arranged in a 4×1 block portion of an 8×8 block. The 16 equations g1(0,0) to g1(1,7) of the above “Interlaced-Scan, Base-layer Set of Decimated Pixel Values” were derived by substituting equation 3 into equation 6, substituting numeric values for x and y in g1′(x,y), substituting N=8, and approximating the weighting factors for the DCT coefficients with rational values. The 16 equations g0(0,0) to g0(2,7) of the above “Interlaced-Scan, Enhancement-layer Set of Decimated Pixel Values” were derived in a similar manner by substituting equation 3 into equation 7, substituting numeric values for x and y in g0′(x,y), substituting N=8, and approximating the weighting factors for the DCT coefficients with rational values. Although the effective pixel decimation of the enhancement-layer is only 2 (rather than the effective pixel decimation of 4 of the base-layer), the equalities g0(x+1,y)=−g0(x,y) hold for x=0 and x=2 so that enhancement-layer values with odd horizontal indexes need not be computed. Thus, only 16 independent g0(x,y) enhancement-layer values need be computed for each 8×8 luma block in an interlaced-scan I or P picture. Further, because these 16 g0(x,y) enhancement-layer values are residual values, they tend to have a small dynamic range.

Returning to FIG. 2, unit 204 conveys an output comprising successive 8×8 blocks of I, P and B luma and chroma g1(x,y) base-layer decimated pixel values as a first input to base-layer adder 205B in a predetermined order. (For non-coded blocks all such values are zero). This predetermined order includes the decimated pixel values of each 2×2 array of 8×8 pixel luma blocks and each of two chroma blocks which form a decimated macroblock for use by enhanced MCU processing means 208. Further, unit 208 applies a corresponding block p1(x,y) of base-layer decimated pixel values as a second input to base-layer adder 205B in this same predetermined order (For intra-coded macroblocks all such values are zero). The block s1(x,y) of base-layer decimated pixel values derived as a sum output from base-layer adder 205B are then stored in memory 206.

Unit 204 conveys an output comprising the I and P luma g0(x,y) enhancement-layer decimated pixel values as a first input to enhancement-layer adder 205E in the previously mentioned decimated-pixel macroblock predetermined order. (For non-coded blocks all such values are zero). Further, for the case of P luma pixels, unit 208 applies a corresponding macroblock of 64 p0(x,y) enhancement-layer decimated pixel values as a second input to adder 205E in this same predetermined order. (For intra-coded macroblocks all such values are zero). The macroblock of 64 s0(x,y) enhancement-layer decimated pixel values derived as a sum output from adder 205E are applied as an input to enhancement-layer encoder 207 and then the encoded output bit-words from encoder 207 are stored in memory 206 during decoding of I and P pictures.

A macroblock at the higher resolution of the enhancement-layer would normally comprise 128 decimated luma pixel values. However, because of the above-described symmetry equalities for both progressive-scan sequences and interlaced-scan sequences, the number of independent decimated enhancement-layer pixel values in the block s0(x,y) is reduced from 128 to 64. Therefore, the predetermined order is such that only half of the enhancement-layer decimated pixel values need be considered by enhancement-layer encoder 207. These enhancement-layer values are encoded in pairs using a simple vector quantizer, with each pair of values being represented by an 8-bit codeword. Since there are 64 enhancement-layer values to be encoded in a macroblock, the number of bits of storage for the enhancement layer is 32×8=256 bits per macroblock. In the preferred embodiment the 32 codewords are combined into two 128-bit output words from encoder 207 for storage in memory 206.

For progressive sequences each pair of horizontally adjacent values in the block s0(x,y) is encoded as a two-dimensional vector, whereas for interlaced sequences each pair of vertically adjacent (within the same field) values in s0(x,y) is encoded as a two-dimensional vector. Let v0 and v1 be a pair of values to be encoded together. The computational procedure employed by encoder 207 to encode the pair v0,v1 is described in detail in Appendix A. After this procedure has been completed for each pair of values in s0(x,y), the codewords are packed into two 128-bit words, both of which 128-bit words form the output from encoder 207 that are stored in memory 206. Returning again to FIG. 2, memory 206 provides (1) a base-layer output d1(x,y) to unit 208 (d1(x,y) is similar in content to the base-layer input s1(x,y) provided to memory 206) and (2) an enhancement-layer output to unit 208 (similar in content to the enhancement-layer input to memory 206).

In order for enhanced MCU processing means 208 to form a block of predictions, a block of pixel values is fetched from memory 206. The base-layer of pixel values which are read from the stored reference picture are denoted d1(x,y). The enhancement-layer residual values, which are needed only if the block of predictions being formed is for the luma component in a P picture, are denoted d0(x,y). Since the enhancement-layer samples are stored in memory 206 in encoded form, the enhancement-layer data output from memory 206 input to unit 208 is decoded by enhancement-layer decoder 300 (FIG. 3) to obtain the d0(x,y) values. Unit 208 separately forms individual luma or chroma outputs for field prediction operations corresponding to the top and bottom field prediction blocks. In a bi-directionally predicted macroblock these operations are performed separately for the forward and backward predictions and the results are combined as described in the ISO 13818-2 standard. In the following detailed description of the computational-processing operations performed by unit 208, the symbol / represents integer division with truncation of the result toward minus infinity, the symbol // represents integer division with truncation of the result toward zero, and the symbol % represents the modulus operator, which is defined such that if x is a negative number and M is a positive number, then x % M=M−((x//M)*M−x).

Before a block of samples can be read from memory 206, the location and size of the block is determined. The location of a block of pixel values in the reference picture is specified by the horizontal and vertical coordinates of the start (i.e., the upper-left corner) of the block in the reference picture. For the base-layer, these coordinates are indexes into a picture which is ¼ horizontal, full vertical resolution for interlaced sequences and ½ horizontal, ½ vertical resolution for progressive sequences. For the enhancement-layer, the coordinates are indexes into a picture which is ½ horizontal, full vertical resolution for both interlaced and progressive sequences.

To locate the blocks d1(x,y) and d0(x,y) in the reference picture, the motion vector for the macroblock being decoded is needed. The decoding of motion vector data in the bitstream, the updating of motion vector predictors, and the selection of motion vectors in non-intra macroblocks which contain no coded motion vectors (e.g., skipped macroblocks) are all performed by unit 208 as described in the ISO 13818-2 standard. Let xb and yb be the full-resolution horizontal and vertical positions of the macroblock being decoded and let mv=(dx,dy) be the decoded motion vector, so that if the sequence were being decoded at full resolution, a block of pixel values at location (xb+(dx/2), yb+(dy/2)) in the full-resolution reference luma picture would be read from memory and used to form luma predictions. Similarly, a block of chroma values at location (xb/2+(dx//2), yb/2+(dy//2)) in the reference chroma picture would be needed to form predictions for each of the 2 chroma components in a full-resolution mode.

The location in the reference picture of a block needed for motion compensation in unit 208 is determined using xb, yb, dx and dy. Table 3, shows the locations of blocks for various prediction modes. The sizes of the blocks needed for motion compensation in unit 208 are specified in Table 4. Base-layer entries in Table 4 give the size of the block d1(x,y), and enhancement-layer entries in Table 4 give the size of the block d0(x,y).

TABLE 3
Locations of Blocks Needed for Motion Compensation
in Enhanced MCU Processing Means 208
Horizontal Vertical
Prediction Mode Coordinate Coordinate
Progressive sequence, luma, ((xb + (dx/2))/8)*4 (yb + (dy/2))/2
base-layer
Progressive sequence, luma, ((xb + (dx/2))/8)*4 ((yb + (dy/2))/2)*2
enhancement-layer
Progressive sequence, chroma xb/4 + ((dx//2)/4) yb/4 + ((dy//2)/4)
Interlaced sequence, luma, ((xb + (dx/2))/8)*2 yb + (dy/2)
base-layer
Interlaced sequence, luma, ((xb + (dx/2))/8)*4 yb + (dy/2)
enhancement-layer
Interlaced sequence, chroma xb/8 + ((dx//2)/8) yb/2 + ((dy//2)/2)

TABLE 4
Sizes of Blocks Needed for Motion Compensation
in Enhanced MCU Processing Means 208
Horizontal Vertical
Prediction Mode Size Size
Progressive sequence, luma, base-layer 12 9
Progressive sequence, luma, enhancement-layer 12 18
Progressive sequence, chroma 5 5
Interlaced sequence, luma, 16 × 16 prediction, 6 17
base-layer
Interlaced sequence, luma, 16 × 8 prediction, 6 9
base-layer
Interlaced sequence, luma, 16 × 16 prediction, 12 17
enhancement-layer
Interlaced sequence, luma, 16 × 8 prediction, 12 9
enhancement-layer
Interlaced sequence, chroma, 8 × 8 prediction 3 9
Interlaced sequence, chroma, 8 × 4 prediction 3 5

FIG. 3 shows the processing performed on the luma samples read from memory 206 by unit 208. As shown in FIG. 3, the luma-processing portion of unit 208 comprises enhancement-layer decoder means 300, enhancement-layer pixel reconstruction means 302, DCT-based upsample means 304, full-resolution block select means 306, DCT-based downsample means 308 and two-layer output formation means 310. These elements of enhanced MCU processing means 208 use the reduced-resolution blocks from I and P frames stored in memory 206 to form predictions for a decoded macroblock.

The above-described structure of FIG. 3 performs computational processing of luma pixel values input to unit 208 from memory 206. This computational process is described in detail in appendices B to G. Briefly, however, decoder 300 unpacks the input 128-bit words into 16 constituent 8-bit codewords. Decoder 300, employing the computational processing described in appendix B, derives d0(x,y) as an output. Enhancement-layer pixel reconstruction means 302, employing the computational processing described in appendix C, derives r0(x,y) as an output in response to both the d1(x,y) input to unit 208 and the d0(x,y) output from decoder 300. DCT-based upsample means 304, employing the computational processing described in appendix D, horizontally upsamples the r0 (x,y) input to derive r(x,y) at full resolution. Full-resolution block select means 306, employing the computational processing described in appendix E, uses r(x,y) to derive a full-resolution block of predictions p(x,y) as an output. DCT-based downsample means 308, employing the computational processing described in appendix F, horizontally downsamples the p(x,y) input to derive q(x,y) at half horizontal resolution. The block q(x,y) is applied as an input to two-layer output formation means 310, which employs the computational processing described in appendix G, to derive the outputs p1(x,y) and p0(x,y) provided by unit 208 to adders 205B and 205E shown in FIG. 2. Although the computational processing required for chroma predictions are not shown in FIG. 3, they are described in detail in appendix H.

Returning again to FIG. 2, a video signal comprising the base-layer pixels defining each of successive picture fields or frames is output from memory 206 and input to sample-rate converter 210 which derives a display video signal output. The display video signal output from unit 210 represents a PIP display image. By way of example, assume that the size of a PIP display is intended to occupy ⅓ of the horizontal dimension and ⅓ of the vertical dimension of the entire HD display size. If the original resolution of the HD bit-stream is 1920×1080 interlaced, then the PIP decoded frames (in which the number of pixels in the horizontal direction has been decimated by a factor of ¾) are 480×1080 interlaced. Assuming a 1920×1080 interlaced HD display, the displayed PIP frames should be 640×360 interlaced. Therefore, in this example, the decoded frames stored in memory must be scaled by sample-rate converter 210 by a factor of 4/3 in the horizontal direction and ⅓ in the vertical direction.

In a realized embodiment of the present invention, the extra capacity required in 1 memory for storage of the ½ resolution enhancement-layer in encoded form adds only 1.98 Mbits to the 17.8 Mbits required for storage of the ¼ resolution base-layer. Thus, the inclusion of an encoded ½ resolution enhancement-layer increases the needed storage capacity of the base and enhancement-layer decimated-pixel memory by a relatively small amount (i.e., only a little more than 11%) to 19.78 Mbits.

Horlander, Thomas Edward, Comer, Mary Lafuze

Patent Priority Assignee Title
10154269, Jan 02 2014 VID SCALE, Inc. Methods, apparatus and systems for scalable video coding with mixed interlace and progressive content
10277898, Feb 29 2012 Sony Corporation Image processing device and method for improving coding efficiency of quantization matrices
10404985, Feb 29 2012 Sony Corporation Image processing device and method for improving coding efficiency of quantization matrices
7734151, Dec 06 2004 LG Electronics Inc Method for decoding image block
8014613, Apr 16 2007 Sharp Kabushiki Kaisha Methods and systems for inter-layer image parameter prediction
8259813, Dec 19 2007 Cisco Technology, Inc Method and system for improving color sharpness in video and images
8391370, Mar 11 2009 Hewlett-Packard Development Company, L.P.; HEWLETT-PACKARD DEVELOPMENT COMPANY, L P Decoding video data
8401091, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
8433184, Dec 06 2004 LG Electronics Inc. Method for decoding image block
8451899, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
8457201, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
8494042, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
8494060, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
8619872, Jan 09 2006 LG Electronics, Inc. Inter-layer prediction method for video signal
8792554, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
9406112, Jul 10 2008 KONINKLIJKE PHILIPS N V Video data compression
9497453, Jan 09 2006 LG Electronics Inc Inter-layer prediction method for video signal
9819947, Jan 02 2014 VID SCALE, INC Methods, apparatus and systems for scalable video coding with mixed interlace and progressive content
9961303, Mar 04 2013 Cisco Technology, Inc. Video conference virtual endpoints
Patent Priority Assignee Title
5227878, Nov 15 1991 MULTIMEDIA PATENT TRUST C O Adaptive coding and decoding of frames and fields of video
5253056, Jul 02 1992 SHINGO LIMITED LIABILITY COMPANY Spatial/frequency hybrid video coding facilitating the derivatives of variable-resolution images
5341318, Mar 14 1990 LSI Logic Corporation System for compression and decompression of video data using discrete cosine transform and coding techniques
5371549, Oct 07 1992 QUARTERHILL INC ; WI-LAN INC Decoding method and system for providing digital television receivers with multipicture display by way of zero masking transform coefficients
5386241, Oct 07 1992 QUARTERHILL INC ; WI-LAN INC Signal transformation system and method for providing picture-in-picture in high definition television receivers
5598222, Apr 18 1995 Hatachi American, Ltd.; Hitachi America, Ltd Method and apparatus for decoding multiple video bitstreams using a common memory
5614952, Oct 11 1994 Hitachi America, Ltd Digital video decoder for decoding digital high definition and/or digital standard definition television signals
5614957, Nov 14 1994 Hitachi America, Ltd. Digital picture-in-picture decoder
5635985, Oct 11 1994 Hitachi America, Ltd.; Hitachi America, Ltd Low cost joint HD/SD television decoder methods and apparatus
5737019, Jan 29 1996 Panasonic Corporation of North America Method and apparatus for changing resolution by direct DCT mapping
5767797, Jun 18 1996 Kabushiki Kaisha Toshiba High definition video decoding using multiple partition decoders
5847771, Aug 14 1996 Verizon Patent and Licensing Inc Digital entertainment terminal providing multiple digital pictures
5867601, Oct 20 1995 Panasonic Corporation of North America Inverse discrete cosine transform processor using parallel processing
EP707426,
WO9841011,
WO9841011,
////
Executed onAssignorAssigneeConveyanceFrameReelDoc
Oct 25 1999HORLANDER, THOMAS EDWARDThomson Consumer Electronics, IncASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0103530382 pdf
Oct 26 1999COMER, MARY LAFUZEThomson Consumer Electronics, IncASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0103530382 pdf
Oct 28 1999Thomson Licensing S.A.(assignment on the face of the patent)
Jan 13 2000Thomson Consumer Electronics, IncTHOMSON LICENSING S A ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0105250971 pdf
Date Maintenance Fee Events
Jun 08 2009M1551: Payment of Maintenance Fee, 4th Year, Large Entity.
Mar 08 2013M1552: Payment of Maintenance Fee, 8th Year, Large Entity.
Jun 09 2017M1553: Payment of Maintenance Fee, 12th Year, Large Entity.


Date Maintenance Schedule
Jan 31 20094 years fee payment window open
Jul 31 20096 months grace period start (w surcharge)
Jan 31 2010patent expiry (for year 4)
Jan 31 20122 years to revive unintentionally abandoned end. (for year 4)
Jan 31 20138 years fee payment window open
Jul 31 20136 months grace period start (w surcharge)
Jan 31 2014patent expiry (for year 8)
Jan 31 20162 years to revive unintentionally abandoned end. (for year 8)
Jan 31 201712 years fee payment window open
Jul 31 20176 months grace period start (w surcharge)
Jan 31 2018patent expiry (for year 12)
Jan 31 20202 years to revive unintentionally abandoned end. (for year 12)